Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2602.00950 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910007613718528 |
|---|---|
| author | Farinhas, António Guerreiro, Nuno M. Pombal, José Martins, Pedro Henrique Melton, Laura Conway, Alex Dochat, Cara D'Eon, Maya Rei, Ricardo |
| author_facet | Farinhas, António Guerreiro, Nuno M. Pombal, José Martins, Pedro Henrique Melton, Laura Conway, Alex Dochat, Cara D'Eon, Maya Rei, Ricardo |
| contents | Large language models are increasingly used for mental health support, yet their conversational coherence alone does not ensure clinical appropriateness. Existing general-purpose safeguards often fail to distinguish between therapeutic disclosures and genuine clinical crises, leading to safety failures. To address this gap, we introduce a clinically grounded risk taxonomy, developed in collaboration with PhD-level psychologists, that identifies actionable harm (e.g., self-harm and harm to others) while preserving space for safe, non-crisis therapeutic content. We release MindGuard-testset, a dataset of real-world multi-turn conversations annotated at the turn level by clinical experts. Using synthetic dialogues generated via a controlled two-agent setup, we train MindGuard, a family of lightweight safety classifiers (with 4B and 8B parameters). Our classifiers reduce false positives at high-recall operating points and, when paired with clinician language models, help achieve lower attack success and harmful engagement rates in adversarial multi-turn interactions compared to general-purpose safeguards. We release all models and human evaluation data. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_00950 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | MindGuard: Guardrail Classifiers for Multi-Turn Mental Health Support Farinhas, António Guerreiro, Nuno M. Pombal, José Martins, Pedro Henrique Melton, Laura Conway, Alex Dochat, Cara D'Eon, Maya Rei, Ricardo Artificial Intelligence Large language models are increasingly used for mental health support, yet their conversational coherence alone does not ensure clinical appropriateness. Existing general-purpose safeguards often fail to distinguish between therapeutic disclosures and genuine clinical crises, leading to safety failures. To address this gap, we introduce a clinically grounded risk taxonomy, developed in collaboration with PhD-level psychologists, that identifies actionable harm (e.g., self-harm and harm to others) while preserving space for safe, non-crisis therapeutic content. We release MindGuard-testset, a dataset of real-world multi-turn conversations annotated at the turn level by clinical experts. Using synthetic dialogues generated via a controlled two-agent setup, we train MindGuard, a family of lightweight safety classifiers (with 4B and 8B parameters). Our classifiers reduce false positives at high-recall operating points and, when paired with clinician language models, help achieve lower attack success and harmful engagement rates in adversarial multi-turn interactions compared to general-purpose safeguards. We release all models and human evaluation data. |
| title | MindGuard: Guardrail Classifiers for Multi-Turn Mental Health Support |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2602.00950 |